fix(ops): default core GCP env on Cloud Run deploy

CI and minimal shell envs omitted GOOGLE_CLOUD_PROJECT and BQ pins, so the
runtime BigQuery tool had no compute project; always set project, Vertex, and dataset when deploying.

Author: samartho4

backend/deployment/deploy.py

@@ -117,7 +117,15 @@ def _commands() -> list[list[str]]:
 
     # repo root holds the Dockerfile + .gcloudignore; --source . builds it via Cloud Build.
     repo_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-    env_pairs = [f"{k}={os.environ[k]}" for k in _RUNTIME_ENV_KEYS if os.environ.get(k)]
+    # Always pin project/Vertex/BQ dataset when PROJECT is set — CI deploy often omits these in the shell env,
+    # which leaves settings.project empty and BigQueryToolset fails with "project not found".
+    runtime: dict[str, str] = {k: os.environ[k] for k in _RUNTIME_ENV_KEYS if os.environ.get(k)}
+    if PROJECT:
+        runtime.setdefault("GOOGLE_CLOUD_PROJECT", PROJECT)
+    runtime.setdefault("GOOGLE_CLOUD_LOCATION", REGION)
+    runtime.setdefault("GOOGLE_GENAI_USE_VERTEXAI", "true")
+    runtime.setdefault("BQ_DATASET", "dry_lab")
+    env_pairs = [f"{k}={runtime[k]}" for k in _RUNTIME_ENV_KEYS if k in runtime]
     secret_pairs = [f"{env}={name}:latest" for env, name in _SECRET_REFS.items()]
     deploy = [
         "gcloud", "run", "deploy", SERVICE,